Something Random

[sedrosken]

You end up with strange issues with older machines. In another effort to get just a little more performance out of my 486 -- I really, really just wanted to see Diablo play smoothly on it, and seeing 32-channel XM modules play smoothly would be nice as well -- I've managed to source a few PGA168 interposers that offer a way to force a 3/4x multiplier for boards without support for that, and voltage regulation to take the 5v the board provides down to the 3.3v that won't fry newer CPUs. I used the Am5x86 I'd previously used with the M912 board, and thankfully it's not fried -- initially I had the sort of success that I never do, everything just worked. CPU-Z made the thing BSOD, but that was actually down to the VxD it generated to get access to the CPUID I suppose -- deleting the old one and letting it remake it solved it.

No, I dipped back into the case and noticed I had some residue on the AA batteries where I'd superglued the holder to the power supply in lieu of using less-permanent mounting options, as the others I'd tried lost their adhesion after a while. Without thinking I removed the batteries to clean this off, and lost all my CMOS settings. When going through and resetting all those settings, I have a few that actually have a reputation for being unstable on this class of hardware; namely, I have everything (cache, RAM, local bus) set to zero wait-states, and I'm both shadowing and caching an option ROM at address D000-D7FF. You can see why I initially thought one of these other settings must have been causing the problems I'll describe in a moment -- they are known to cause instability on other 486-class machines, and while they worked reliably here before, maybe something changed with the upgrade I did.

When booting, particularly in my normal configuration loading EMM386 and then initializing my sound card at address 220h, interrupt 5, low DMA 1, and high DMA 5, with the MIDI port at 330h, the machine would totally freak out, drop to 40-column text mode, and then display an error message that seemed pretty unsure of itself:

"MEMORY PARITY ERROR ????"

It turns out that I'd inadvertently toggled the Gate A20 Emulation setting from Chipset to Fast. For those unfamiliar, this dictates how expanded memory is accessed, and as most 386+ DOS configs make use of EMM386 to provide expanded memory emulation using extended memory for applications requiring it, it's a surprisingly important setting. On later machines I'm sure this works fine in Fast mode, indeed I remember it causing no issues back when I had the Pentium Pro machine. However I'm guessing that the UM480, being a combination 386/486 chipset made at a time when some people still used genuine expanded memory cards, might just be early enough that it's got some weird bugs with that mode of operation.

One CMOS setting had the 486 out of commission for weeks between me not having time to diagnose further and simply having no idea what could possibly be wrong. Imagine my terror when swapping back to the DX4 OverDrive, a known-good CPU, didn't fix the problem! I reseated and used deoxit on the contacts of all my RAM and every add-in card, several times, swapped hardware around, and only when I decided to try loading the BIOS defaults on a whim did I fix it purely by accident.

Ironically this is still not the final chapter -- I want to find a specific model of AMD DX4 CPU, one rated for 120MHz operation (so I can use the 40MHz crystal and run a 40MHz system bus) but that only has 8K of L1 cache, as I'm almost completely certain that's what's tripping up the chipset as far as enabling L2 cache -- these later chips all have 16K. When I finally find one of this specific model that doesn't come from China, doesn't cost two hundred dollars, or perhaps both, I'll have one last thing to test. Thankfully I do have the 12ns cache chips floating around from my M912 misadventure, so if/when I get that working I'll have faster cache for 0WS operation at 40MHz.

 

[Mercutio]

You might have a better time if you just tracked down an OG 486DX/50 than messing with the 3rd party clock multiplied CPUs. Cyrix and AMD both made absolute garbage in that era, and the reason the DX/50 was so badass that everything else clocked to 50MHz along with the CPU. You'll probably get the best graphics off whatever VLB nonsense you have (ET6000? S3 Trio?) from the bus speed improvement over the extra CPU clock.

 

[sedrosken]

My board offers support for a 50MHz bus, but none of the cards I have will work at that speed -- even 40MHz is technically overclocking it. And in benchmarks, the DX2-66 beats the DX-50 in roughly 75% of cases, particularly CPU-bound ones, while a DX4-120 beats it in all possible scenarios -- mostly because the 50MHz machine has ISA video since VLB cards that work at 50MHz are so few and far between that realistically the ones that do wouldn't be paired with a DX1 anyway. Because of the way VLB works -- essentially mapping the device and its associated memory directly into the 486 processor's memory map, and with the 486 not really being designed to directly drive devices like this, the generally accepted order of operations is as follows:

- At 50MHz, you'll get one card to work if you're lucky -- usually you'll need to throw a waitstate or two at it.
- At 40MHz, with careful selection and maybe a waitstate for good measure you'll get two cards working -- I just so happen to have two cards that are known to work fine together at that clock, with no waitstates, even! I have a Trio64V+ and a UM8672-based IDE controller. (You're probably thinking of the ET4000/W32p, Merc, the ET6000 was a purely PCI card.)
- At 33MHz, you tend to be able to use all three slots with no waitstates. This is the "design" frequency of the bus, it's what most cards were made to target. Seeing how most people only ever went up to a DX2-66 and then directly onward to a Pentium or some other such, that's about right. Remember, VLB was only supposed to fill the gap until PCI was finished up and ready to go -- that companies continued to make cards following the spec even into 1995 was something of a miracle.
- At 25MHz, the bus is technically underclocked, so if your hardware doesn't even work at this bus speed, it's time to consider other boards.

As for the AMD/Cyrix parts being inferior -- in the Pentium era, that's largely true, and the Cyrix 486s are a bit behind particularly in floating point here, but the AMD stuff is pretty much identical. To the point they got sued for using some Intel microcode, even. The main difference is that outside the DX/SX-40, Intel never bothered making another 40MHz bus part again because of the issues they ran into, and the DX-50 was a rare bird indeed. Usually only found in systems crippled with ISA-only motherboards. AMD and Cyrix both tried their hand at a DX2-80, and again ran into issues mostly stemming from VLB instability, by this point the boards were plenty capable of going that fast. With the DX4-120 that AMD put out, the issues largely disappeared because most people used these on PCI motherboards or newer VLB boards that could cope with the overclock. The Am5x86 only ever officially ran at 4x33 for 133MHz, in parts that people actually got their hands on anyway, but they are well documented to generally be capable of 3x50 for 150 and 4x40 for 160MHz respectively, some models being better for this than others. A couple brave souls drove the record for the 486 platform up to 200MHz by running at 4x50. Interestingly the Cyrix 5x86 part was actually a cut-down 6x86, and it was easily the most advanced (and thus most incompatible and fussy) chip for the socket as a result. Generally it scores pretty disappointingly in benchmarks, even with all the special features enabled it is still generally outrun by a Pentium OverDrive. In actual use I'm told it's actually about as good as its big brother under Windows 9x.

 

[LunarMist]

The 486 DX-50 was not rare as far as I recall. There was a small room full of them at one site in the early 90s. Heavy lifting was done by a couple of Alpha (DEC) systems that the likes of me were not allowed to touch. I had to use a 386. Nobody in MIS would buy any x86 systems that were not Intel. The off-brands were mostly for cheap home systems or quite small businesses.

 

[sdbardwick]

The 486 DX-50 was not rare as far as I recall. There was a small room full of them at one site in the early 90s. Heavy lifting was done by a couple of Alpha (DEC) systems that the likes of me were not allowed to touch. I had to use a 386. Nobody in MIS would buy any x86 systems that were not Intel. The off-brands were mostly for cheap home systems or quite small businesses.

DX-50s were rather rare. Fussy with motherboards due to clock and being quickly followed by DX2-66s that worked well with many boards. I think I still have mine around somewhere, along with some AMD DX4s and Cyrix chips (just kept the CPUs for fun, didn't keep the bulky components).

Tannin agrees with rarity:

A very rare bird indeed. In theory, the DX-50 should have been a winner, with 50MHz processing power and a 50MHz bus too, but in practice it was just too fast for motherboards to cope with. We sold only one of them, and had no end of trouble with it. In the end, we gave up changing motherboards yet again and clocked it back to 40MHz, making up the speed difference by going to the (then new) VESA local bus video and I/O cards. (40MHz and VESA — and there is another batch of epic struggles to recount — but this particular one was kind to us.)
Actually, VESA had a lot to do with the market failure of the DX-50. It was hard to make an ISA system run reliably at 50MHz, and almost impossible with VESA. PCI wasn't out then, and when it did come out it was dreadfully unreliable for the first year or so. By the time of the Pentium 75, 50MHz and even 66MHz bus speed was easily achievable: in the time of the 486DX-50 though, 40MHz was tricky stuff and 50MHz boards were a nightmare

 

[sedrosken]

I hoped someone would pick up on the Red Hill reference I made.

 

[Mercutio]

Yes I misremembered the ET6000. I have it in my head that Tseng Labs didn't survive into the PCI era.

The rarity might've been an Aussie issue though. I definitely built a few systems using DX/50s. I was in my late teens then, but I had no trouble ordering or assembling systems with them. Beats me what board I used. Might've been Octek or Zeos or Micronics at the time. My bread and butter were the DX/33s, and I recall being very happy with 2MB ET4000 video cards (they allowed for 1280x1024 32bpp in Windows or OS/2) and upselling 17" monitors, which were just starting to be affordable in 1993. I remember getting wholesale Mag Innovision DX17Fs for something like $550 at a point in time that 14" screens were still going for around $150, but the difference was big enough that I could find people who would pay for it, since big box stores weren't even offering "large" displays at the time. Home computing as a weird thing in the early 90s, with a lot of the focus being on getting a Multimedia PC that had both a sound card and a CD-ROM drive. For a while, the NEC drives I was buying actually included a bundle of CDs and a 40 minute VHS tape that explained how to install one! I still have a box from one around somewhere. I remember sound cards as the biggest pain in my ass for a couple years, and being pretty insistent that anyone who wanted a modem get an external one so I wouldn't have to deal with COM3/4 headaches.

From what I remember, both Cyrix and AMD were notorious for getting hot, although Cyrix was by reputation the worst of it.

 

[sedrosken]

Yes I misremembered the ET6000. I have it in my head that Tseng Labs didn't survive into the PCI era.

Yeah, the ET4000/W32p was their first PCI-capable chipset, though it apparently wasn't particularly fast on PCI like it was on VLB. The ET6000 was very nice for the time, but given the choice we have today, I'd take a Matrox card over an ET6000 every time. The Matrox will be faster, have more stable drivers, and have a much better output quality too. For some reason people don't seem to be hoarding Matrox cards like they are nVidia/ATi stuff so you can actually find one for a decent price these days.

From what I remember, both Cyrix and AMD were notorious for getting hot, although Cyrix was by reputation the worst of it.

That may very well be the case, but I go overboard with my cooling anyway. I have a heatsink that wouldn't be out of place for a mid-level Socket 7 machine installed on my Am5x86, not because it particularly needs it (the thing barely gets warm to the touch), but because I'm paranoid and would like this thing to last ideally as long as I can get it to. I know the Cyrix ones are very hot indeed, because the manufacturing facilities they had access to at the time were, compared to those of AMD and especially Intel, quite primitive. Frankly, if their deal with IBM didn't go through, they probably never would have been able to make the 6x86, and even that had heat problems.

This week in Impulse Purchases™, I snagged a Japanese New 3DS LL (their version of the XL since X isn't used to denote size there) in pretty decent shape. It's going to take a month or so to get here because customs, but once it gets here it's off to the races. I bought the Japanese version because that's easily the cheapest to acquire, the American version starts at roughly twice the price. I can't read a word of Japanese but apparently if you softmod it (and there's a universal exploit that works on every region, every firmware) you only need to blindly navigate the katakana maze once, since the homebrew and all the launchers tend to be in English and the process makes the console region-free. As for why I did this, I've been wanting a 2DS or 3DS since I lost my last one in my breakup with my ex. I didn't have many games for it but I loved what I did have. Plus, the DS backcompat is pretty much perfect aside from the Guitar Hero games that use the grip that attached to the GBA slot, and there's apparently a way to hijack the Virtual Console emulator to run any ROM you want from the consoles it supports, to not even get into the homebrew scene.

 

[Mercutio]

It Could Happen Here is a legitimately interesting podcast. The oldest episodes in the feed are about the ramifications of civil war and the Balkanization of the USA, from a journalist who spent his career in conflict zones like Syria and Georgia and from relevant subject matter experts.
The podcast recently relaunched with an updated focus on climate and the crumbling infrastructure we seem to have in the USA and western world.
Today, the podcast's guest was Chelsea Manning, who discussed practical matters for individual information security. It's worth a listen.

 

[Mercutio]

Shang Chi wasn't what I thought it would be. Marvel made a Wuxia movie, akin to Crouching Tiger, Hidden Dragon, by way of both frequent nods to Chinese mythology and the now clearly defined Marvel Cinematic Universe. In comics, Shang Chi was absolutely created to tie in to Bruce Lee and Shaw Brothers in their biggest moment, but rather than play on any of that, we have another sympathetic villain, strong comic relief and some fantastic fight choreography. I really like the way this movie handles color. It's very clear who is who throughout and the action is all the more impressive for how easy it is to follow.

If you have small people to take, yes, it's violent in terms of punching and kicking, but there's no graphic bloodshed.

 

[sedrosken]

A bit random, but that's what this topic's for...

What VPS provider would you recommend? I don't want or need anything spectacular, though I could be persuaded to splurge a little if they offer free or cheap domain registration as a bonus. My budget is, like, $30/mo maximum, but that's if it does everything I want -- let me explain.

Bare minimum essential needs are the ability to run ZNC (an IRC bouncer), and an instance of Apache2. I don't expect ridiculous amounts of traffic, in fact the site will be pretty barebones -- HTTP only and HTML3 compliant at best. That can run on any old potato, heck if my ISP didn't go ballistic on people hosting from residential connections I'd consider self-hosting again. I'd prefer no bandwidth cap, just to not have to think about it. The most I'd spend on this basic of a server would be, like, $10/mo.

Nice-to-haves would be 4c/8GB RAM, as then I'd have some breathing room to use on a MariaDB server and a couple game servers, probably just Quake III and an Asheron's Call server. I could probably squeak it by on 4GB/2c, especially since I expect to be the only player, and it'll be console-only anyway, but I'd prefer to not have to do that.

Storage is not a concern. I won't be running a file dump, the biggest space sink will be the logs from ZNC I expect. I've got that running from a free-tier AWS server with an 8GB virtual hard disk and it's barely sitting at half-used all told. As for why I'm unhappy with Amazon, I mostly just don't want to give them any more data than I have to, and I've heard some rumblings that the free tier only lasts about a year or so.

 

[sedrosken]

You can disregard this last post. I ended up looking on LowEndBox for a deal and found one with RackNerd -- for 24 dingleberries a year I have a VPS with 2.5GB RAM and 2 cores on an E5-2680v2, with 40GB of space to play with. Plenty for the base usecase I had in mind, we'll see if I can make it work for MariaDB and that AC server. I may just be able to squeak it. I bought a domain on namecheap for all of $1.18 for the entire year. Entire setup, start to finish, with me slamming my face into a brick wall a few times for good measure, took around four hours.

 

[Mercutio]

FWIW you should be able to get an AWS instance in that range, but if you can find enough to do with a low-end system like that, you can PROBABLY find enough to move on up to a self-hosted box somewhere. It's not hard to make a VM host on a 1U rack a profit-making item.

 

[sedrosken]

Yeah, I was considering repurposing another Sandy Bridge-era Optiplex for that, maybe install a RAID card and drop a few drives in it for RAID5 and use it for a home NAS while also running my website and bouncer from that, but supposedly my new ISP is way stingier about people putting stuff on port 80. Which, I guess, fair enough -- I didn't pay for the commercial connection, after all. Besides, my connection goes down once or twice a night, while it's fine for residential use it's really not ideal for something that's supposed to be up 24/7/365.

The database and game servers would be better used on such a local server as well, but for now, it works, so I'm not going to change anything while I'm still going through some significant changes.

 

[Handruin]

Check out Linode for a cheap VPS? That's what's been hosting this site for many years.

 

[sedrosken]

I actually ended up with something from RackNerd, they had a special cooperative deal with LowEndBox for a 2.5GB/3c/40GB VPS that I jumped on for I think 24 bucks a year? It's more than adequate for what I need.

 

[sedrosken]

Swapped my 486 motherboard/CPU/RAM with a friend of mine for their Super Socket 7 board/CPU/RAM. The 486 is no more, replaced by a machine running a K6-III+. This machine also replaces my Pentium III since it does well enough at both DOS and Win9x to meet my needs. Anything it can't do that the PIII could is better off on my P4 or main computer anyway.

It also helps that I robbed the PIII of its Voodoo3 for this machine, as it's an ideal card for the platform's iffy AGP support. It's well known that VIA didn't support AGP particularly well, and this is an early, early early implementation to boot, frankly I'm surprised it supports 2x mode. Later AGP Voodoos were sometimes lambasted but always notable for their lackluster use of the bus -- mostly they just used it as a 64-bit, 66MHz PCI slot. No AGP texturing, no side-band shenanigans AFAIK.

I initially wanted to go for a dual sound card setup, using my CT2800 for DOS and my Aureal Vortex2 for Windows. I elected to shelve that idea because of the fact that this board has five total usable slots -- one AGP, 3 PCI (last one is a combination with the first ISA) and 2 ISA, and it was going to be a tight fit getting everything I wanted in here as it is. I ended up having the spare slot, but decided to leave the top PCI slot empty to help the Voodoo3's cooling some, also I was already running into IRQ issues as-is and didn't need another to contend with.

A side note -- does anyone know precisely why BIOS makers never allowed end-users to just assign IRQs by slot or something? Hide it in an advanced menu if you must, but I would have been so much better served had such an option existed rather than rolling the dice and saying a prayer every time I reset the config on my ESCD. All I was allowed to do was "reserve" IRQs for legacy ISA devices, and frankly I was shocked that the machine let me bring up my PnP ISA sound card on an IRQ marked "reserved" but I didn't question it.

So then I was going to use my YMF744 as it supposedly has shockingly working DOS support for a PCI card, even the old OPL3 core just sitting usually unused, as well as pretty good 3D sound support. It's no Vortex2, nor is it an Audigy, but for Win9x it has serviceable EAX and A3D acceleration. I ran head-first into a wall as the DDMA option for DOS sound just did NOT work whatsoever. This board is from the time period that SB-link was a thing, but it doesn't provide it, and the card doesn't have it populated anyway -- that said, it's just the work of installing a header and a few simple passives that would fix that.

I then considered my CT2800 on its own, and the Vortex2 on its own in turn, and both left a bad taste in my mouth. The Vortex2 has workable DOS support via emulation through a TSR, but it sounds rather particularly awful and isn't ideal for what is my sole current DOS machine -- with sound, anyway. The CT2800 is rough around the edges, sure it comes up fine with DIAGNOSE and it's got a real OPL3, but it's got a very noisy output and, having the 4.13 DSP, is afflicted with the accursed hanging notes bug. For those who don't know, particularly in Doom-engine and Build-engine games but appearing just about everywhere, Sound Blaster 16 cards with DSP versions greater than 4.05 on up to AWE32s (the AWE64 fixed it for good, but unfortunately only ever came with CQM for FM synth) produce sometimes garbage or illegal notes when the card is used for General MIDI output under DOS. Since the card never intended to send those notes in the first place, it never turns them off so they just sit there... hanging. Until you turn the synthesizer off. It stinks and I didn't want to contend with it here, though notably my particular card is only mildly bugged.

So I revisited the idea of my Audician32, the ISA YMF719-S card. This never worked right in the 486, which I initially bought it for, but apparently it works perfectly fine here, reinforcing my idea that no matter what clock I ran the bus at in that machine, there was always something ever so slightly messed up there. Or perhaps it just needed something capable of executing i586 instructions to work properly, doubtful but who knows. With the drivers off the CD (or downloaded from Yamaha, but critically not the ones bundled with Windows 98SE) you get a nice little XGlite soft-synth that sounds pretty good at the expense of costing CPU/RAM to use. Probably not going to use it often with the, y'know, actual wavetable synth I've got hooked up to it, but having the option is always nice! Setting the port to 388h rather than 220h as it defaults to in SYSTEM.INI, you can use the Voyetra SuperSAPI! 4op FM drivers intended for Windows 3.x with this card, as well. It's... quirky. Very clearly trades polyphony for instrument definition, and almost no MIDIs are made to use it and thus sound like warbling farts. I actually quite prefer the standard FM driver, usually, but there are a few MIDIs that sound absolutely magical with it, so I keep it installed for them and also to meme around on occasion. No 3D sound whatsoever, but on a CPU this slow that's probably actually more of a blessing than anything, especially when you consider the following...

Initially I had some trouble with, for lack of a better word, judder in games like Unreal Tournament, even at resolutions and detail settings that should be cake to this hardware. By the raw numbers, it's running great, averaging around 30fps in a stressful benchmark, sometimes registering up to 50fps. But it would just feel rather stuttery... I first reverted from the AmigaMerlin 3rd party drivers, version 2.9, to the latest 3dfx reference 1.07.00, but that did it even worse it felt like. Out of desperation I reverted further to the earliest DirectX7 driver for the card, 1.03.00. It sounds way earlier, and I suppose it might be, as it uses a wildly different configuration interface, but there's only about a year's worth of development between them. This solved the issue pretty much instantly -- my only workable guess is that somehow these earlier drivers are better optimized for this CPU, whether that be through leveraging 3dnow! extensions better or just plain being better optimized for in-order CPUs.

One final issue I still haven't quite solved, but have alleviated to the point I can live with it if I must, is network throughput. The initial choice for this machine, a 3com 3C-905B, performed so poorly I suspected instability due to my overclock -- this is a 400MHz CPU running at 550, everything I read online tells me the Sharptooth core pretty much universally is capable of this at stock voltage -- even after delving into the DOS-based configuration utility for the card, nothing helped. So I swapped cards, to a slightly newer 3C-905C. This brought the throughput up from a completely unusable 150KB/s to a livable but not ideal 4MB/s. My SD to IDE adapter is fully capable of saturating the UDMA-33 controller VIA provided for this chipset, or is that the other way around, it runs the same here as it did on my Promise Ultra66 or the UDMA-100 controller that was part of the i815. I'm wondering if this is a card issue, an issue between the interaction of the card and the rest of my machine, or something else. 4MB/s is barely a third of the 100mbps this card is rated for, and my server is connected via gigabit so that shouldn't be a problem. Should I see about trying to find a card from a different, albeit still reputable vendor such as Intel? I doubt finding a good PRO/100 card will be that hard or expensive... Realtek is an option but their chipsets abuse the CPU a fair bit more so I'd prefer to avoid them if possible.

 

[Mercutio]

From what I remember, the Via chipsets for SS7 were a step up from the SiS and ALI, but they were all a bit underdone in some way or other. I'm sure I have some Pro/100s sitting around if you want to try one. Probably an AWE64 as well, although for my money the AWE32 was neck and neck with the GUS if you cared at all about MIDI implementation. Have you given any consideration to the possibility of going SCSI rather than messing around with extra IDE controllers?

I had to clean out my office last week, since someone made the terrible decision to replace the carpet. My office is a storage/server room that's always about 10F warmer than the rest of the building and filled with wire rack shelving and bins full of parts. I was rather forcefully asked to get rid of a bunch of stuff that you probably would've killed to dig through, sed, including a collection of new in box vintage motherboards from the mid-90s up to socket AM2.

 

[sedrosken]

From what I've been told, VIA stuff was the more stable offering for SS7, while ALi performed better. Or maybe vice-versa. SiS was the clear budget contender with their integrated graphics baked in, but they were okay too. But I can't say I've had any significant stability problems thus far, it's just got some... personality to it. Like I said, it helps that the Voodoo cards are notorious for not making good use of the AGP bus.

Unfortunately the AWE32s all had the same hanging notes bug, and if by some miracle there was one without it, it'd be one of the later ones using a DSP from an AWE64 but with CQM for FM synth. Yuck. No, I think I've found the sweet spot, but then again I was never really wowed by AWE synth when I have real wavetable stuff to use.

I do have a SCSI card in there, but it's just a 25-pin external port card originally meant for a scanner that I use for some external drives. Internal SCSI stuff is hard to get ahold of these days and witheringly expensive when you do. SCSI2SDs start on the low end for older versions around a hundred bucks. The SD to IDE adapters, since they use readily available SD to CF chips, are something like ten dollars at most, and they perform well enough -- significantly better than period hard drives at around 25MB/s with no seek times. I'm not using any add-in IDE card, this is just the UDMA-33 controller included as part of the MVP3 chipset. I was shocked that apparently VIA made a better UDMA-33 controller than Intel, as this actually does manage to saturate the bus with what my card adapter can handle. A bonus is that this board is apparently late enough that it has no issues with drives under 128GB, though I haven't been able to test with anything more to see if it's got 48-bit LBA issues. That said, 128GB is plenty for a Win98 install with even hundreds of games. I've got a significant chunk of them installed right now, and all my DOS stuff, and it's only sitting around 14-15GB used, which I know would be astronomical for a real 9x PC of the era.

If you find any PRO/100s laying around I'd be glad to take them off your hands, but don't go to any extra effort on my account. I'm sure I would have killed to go through that stuff, but I'm hoping someone managed to rescue it from wherever you had to take it. Any really old stuff has probably cooked itself in its box anyway because pre-1994/5 AT boards typically had NiCad CMOS batteries that like to leak some 20-30 years later.

 

[Mercutio]

If you have any interest in seeing Dune, it's worth seeing in a theatre. It leaked online yesterday and I watched it, but just like Blade Runner 2049, the lingering atmospheric shots and gorgeous lighting deserve a big screen.

I've been to the movie theater five times since May and I haven't sat in audience larger than a dozen people. Granted I go at odd times but I have no idea how theatres can afford to operate.

 

[Handruin]

I'm planning to watch it on Friday when HBO Max releases it. I have the 1080P copy but I'd rather wait for the 4K release with HDR, Atmos, etc to get a better experience. I haven't been to a theater in probably 2 years or more.

 

[sedrosken]

Last night was my last at the restaurant. I'm going full-time with my IT job. No... career.
It feels awfully weird to say that at 23 with no formal education beyond a few basic certs. They believe in me a lot more than I do, I can tell you that much.

 

[Handruin]

I have no undergrad in IT or any certs to my name and I'm also in IT since about age 25. It's probably more common than you might realize.

 

[Mercutio]

I've been entirely ignoring my CS degree since about six months after I graduated. My best friend finished her CS degree in May. She did one three month contract for Amazon and now she's back to dancing.

Don't knock the certs, though. They do get you noticed.

 

[Mercutio]

Today, a porch pirate hit the goddamned "What am I going to do with this?" lottery with my package of 16 4TB U.2 SSDs.
Amazon didn't even mark the package signature required, and the delivery guy just left it by the front door to my customer's office, which is across the street from a 7-11.

Keep a watch on Chicago-area Craigslist and FB Marketplace. Y'all might get a sweet deal on some enterprise drives.

 

[Handruin]

Definitely wasn't knocking certs, just that not having them isn't uncommon in IT and still getting into it as a career. When obtained strategically to align with an interest they can be really useful for career advancements.

 

[Handruin]

Damn, that SSD theft sounds expensive.

 

[Mercutio]

Damn, that SSD theft sounds expensive.

A touch over $16,800 for a box slightly larger than a shoebox. The hilarious thing to me is that Amazon was just like "Yeah, we'll get you 16 more but it'll be Saturday if that's OK." They don't even care.

I kind of want to go dumpster diving in that area tomorrow just to see if I can find them.

 

[Newtun]

So Amazon did accept responsibility (and replacement cost) for failing to "mark the package signature required"?

 

[Mercutio]

So Amazon did accept responsibility (and replacement cost) for failing to "mark the package signature required"?

They just replaced the box. Meaning they sent it out again and didn't charge me for it the second time. The second time, the driver definitely did get a signature on the delivery.

 

[Mercutio]

I completely hate Adobe, but holy crap the neural filters that have recently been added to Photoshop are neat.
Stuff like "copy the makeup off a face and put it on a different face" or "make somebody's stupid face look like a normal, realistic smile" or "completely (re)color a photo."

I know it's all the result of an AI model that's been fed probably petabytes of image data. I'm impressed regardless.

 

[sedrosken]

Possibly a question none of you can answer, but I haven't found a definitive one anywhere: does ERC in a hard drive matter for a Linux soft-RAID via mdadm? The NAS I was allowed to take from work needs new drives and I'm debating whether the ERC of the WD Reds I'm considering over just a fleet of Blues for a RAID 5 is worth it -- namely, if it's there, will mdadm even make use of it? Does it matter? Or will I have a never-ending stream of drives getting kicked from the array for no reason without it? As this NAS is likely to be my main spot for backups (though I will, of course, every so often perform a backup to cold storage) and for my documents and music and such to live, I'd like it to be reliable.

 

[Mercutio]

Why not use drives from a good brand?

 

[sedrosken]

In my personal experience, I've had lots more trouble out of Seagate drives than WD. I know you don't think much of them, but with the consolidation of hard drive makers over the years, it's really coming down to WD vs Seagate, and I don't think much of Seagate anymore. Plus, I'm not sure what of theirs outside their IronWolf range perhaps might even have the ERC support that I may (or may not) need.

Heck -- personal anecdote time -- most of the older Barracuda drives I come across are either dead or dying. My cold storage backup drives are WD Greens that, according to everyone I ask, should have died years ago but are still going strong, albeit a bit slow.

 

[LunarMist]

NAS and Enterprise drives have more specific requirements and a longer time before a timeout is identified that can break the array, if that is what you mean. Years ago consumer WD drives were infamous for (deliberately) performing poorly in arrays to help WD sell their more expensive enterprise drives. Commercial NAS units are more tolerant of drives these days and the FreeNAS is too in my limited experience with that mess. The old WD RED 8TB He drives were decent, but many or all of the newer REDs have the shingles.

 

[sedrosken]

Mostly I just wanted to know if those longer timeouts and such mattered in a mdadm soft RAID, as that's what I'm going to be using them in. I have a semi-reliable source for the pre-shingles WD Reds, as at the capacity I want you can pretty easily get either-or, so SMR isn't an issue.

 

[sedrosken]

So I've been operating under flawed information in regards to my K6-III+ build. I was under the impression that while the addition of on-die L2 cache greatly mitigates the uncached RAM penalty for going beyond what your chipset can cache, it does not completely erase it on the MVP3 platform.

However, according to the benches I've performed on a MVP3 board with 512K on-board L2, which should only be able to cache 128MB, 128MB and 256MB perform identically in the event that an application isn't RAM bound, and of course 256MB performs significantly better in the case that it is.

This greatly alleviates the guilt I felt at potentially hobbling this thing's performance by running 2K on there, as that performs much better in games with 256MB vs 128.

 

[Mercutio]

In my personal experience, I've had lots more trouble out of Seagate drives than WD. I know you don't think much of them, but with the consolidation of hard drive makers over the years, it's really coming down to WD vs Seagate

There are the Toshiba nee HGST drives.
I buy and de-shell a lot of Seagate external drives because I've historically had high demand for bulk storage, but I've honestly not had issues using them in storage arrays, at least not from Windows or my Drobo 8N. I *just* got the 64GB worth of DIMMs I'm about to throw in the PC I wanted to use for testing ZFS. I imagine I'll be looking at that with some of my retired 5 and 6TB SMR drives sometime over the holidays.

 

[Mercutio]

If you still play PC games, Epic is doing a promotion where they apply a $10 off voucher to every game in your cart that costs more than $15 right now.

I've never bought a game from Epic but that is certainly tempting, but I don't really know what its stance is toward DRM.

 

[sedrosken]

EGS is just Steam with shadier business practices, at least as far as DRM is concerned. I refuse to give them my money and I'm one of those obnoxious ones who think you shouldn't either.